[Paw] Towards a PAW WG

["Pascal Thubert (pthubert)" <[e-mail]>]

Dear all:

At the PAW bar BOF in Bangkok, we found common interests in discussing matters similar to DetNet but with a focus on wireless. 

What’s so special with wireless? Well, in very short, scheduling brings all the benefits it does on wires, and more. Scheduling can enable multi channel operations, energy savings, G-MPLS forwarding operations with schedule-based Replication (overhearing), Elimination and Reordering.

But the methods to build a Predictable and Available Wireless network will differ from those used by DetNet/TSN on wires because the medium can be shared, lossy, and broadcast.

This is why we also discussed the need of a group that would specifically address the PAW problem. What would such a group do? 

Quite simply PAW would aim at providing latency and throughput guarantees over a set of radios to be defined, and collaborate with the groups in charge of these radios. In more details:

- Which radios? There are 2 elephants in that room, the adult, ieee std 802.15.4 TSCH for which we have RFC 7554, and the young, ieee sig 802.11 RTA, for which we’d need an informational doc explaining what’s special and interesting about it. Also I’m hearing through the grapevine that the German DLR is developing a new radio that could become a standard for plane communication.

- there is a need to extend the DetNet use cases draft with a wireless twist. Cover real time gaming, in cars and in planes automation, conferencing, building automation, and industrial process control and factory automation.

- there is work that is discussed at BIER and ROLL about methods that could provide extensive time and space diversity using synchronized overhearing and elimination, as well as OAM techniques to optimize the use of the possibly extensive replication that can be needed to provide high delivery ratio with bounded latency. This work needs a central hosting.

- 6TiSCH provides IP best effort over a PAW-capable radio, but did not work on the deterministic flows. We are missing the data model to control the GMPLS forwarding over complex Tracks that is described in the 6TiSCH architecture. PAW would be the natural place to do that.

- ... <anything I missed?>

All and all, there is quite a bit of work ahead. But then, we need to show that we can build a critical mass of people to form a WG. This is where this list comes into play. 

So please respond, indicate what radio we could work on, what network layer problem we need to solve, what amount of energy you are willing to make this happen. Don’t be shy!!

Many thanks in advance...

Pascal